MatrixGen

Overview

• MatrixGen is a modular Python package for generating, resizing, and analyzing sparse matrices. It supports multiple synthesis methods and provides tools for structure-aware scaling, useful in scientific computing, numerical simulations, and ML workloads where realistic matrix patterns are essential.

Features

• Modular Sparse Matrix Generation Generate structurally realistic sparse matrices of arbitrary size from a given base matrix using a unified and extensible framework.

• Multiple Scaling Techniques Includes support for adapted resizing methods such as:

  • Nearest Neighbor

  • Bilinear Interpolation

  • Lanczos Resampling

  • Discrete Fourier Transform (DFT)

  • Discrete Cosine Transform (DCT)

  • Wavelet Transforms

• Preservation of Structure and Sparsity Each method is designed to maintain key structural properties including sparsity patterns, symmetry, and bandwidth.

• Controlled Randomization Allows slight variations during generation to support realistic data augmentation while retaining essential characteristics.

• Feature-Based Evaluation Measures similarity between original and synthetic matrices using structural features like:

  • Nonzero density

  • Bandwidth and profile

  • Symmetry (Psym)

  • Diagonal spread and entropy

  • Row/column distribution metrics

  Cosine similarity is used to quantify structural preservation.

• Scalable Generation Capable of generating matrices at much larger scales than the input while preserving core traits — useful for performance benchmarking and large-scale ML testing.

• Fills a Sparse Matrix Data Gap Addresses the lack of diverse, realistic sparse matrices for algorithm validation, benchmarking, and ML applications.

Requirements

• Handled automatically via Poetry. Requires Python 3.11 or higher.

Installation

You can install MatrixGen via pip from PyPI:

$ pip install matrixgen

Usage

You can use MatrixGen via:

• Python API: from matrixgen import RESIZE_METHODS, resize_matrix, load_matrix, save_matrix
• Command Line: Run matrixgen --help to see options
• See the examples/ folder for usage patterns.

Contributing

Contributions are very welcome. To learn more, see the Contributor Guide.

License

Distributed under the terms of the Apache 2.0 license, MatrixGen is free and open source software.

Issues

If you encounter any problems, please file an issue along with a detailed description.

Credits

This project was generated from @cjolowicz's Hypermodern Python Cookiecutter template.

This project is based off of MatrixGen-d which in itself is a fork of the original MatrixGen by Ali Emre Pamuk, Mert Altekin, and Faruk Kaplan. They are the original creators of the matrix synthesis/generation/resizing logic and without them, this project wouldn't exist.